1. Routing Protocols to Maximize Battery Efficiency
R.R. Rao
University of California, San Diego
C.F. Chiasserini
Politecnico di Torino

2. Outline Life-time of wireless ad-hoc networks
Battery behavior: recovery and rate-capacity effects
The BEE (Battery Energy Efficient) routing protocol
Simulation results

3. Life-time of Ad-Hoc Networks
Network of largely battery-powered nodes with dynamic topology
Objective: Maximize
network life-time; i.e., the time elapsed
until the first node in the network runs out of battery (Tassiulas, Infocom 2000)
S=source; D=destination; R=relay

4. Battery Behavior
Recovery Effect: Battery discharged for short time intervals followed by idle periods (pulsed discharge) is able to deliver a greater energy than under constant current discharge. During idle times battery recovers the charge lost while delivering current pulses
Rate-capacity effect: if the requested current exceeds the rated current specification of the battery, the battery delivers a smaller amount of energy

5. Electrochemical Model of a Li-ion cell
Results obtained by using a program developed by Newman et al. (UC Berkeley)
Time discretized into time intervals 1 ms long
Bernoulli driven discharge: At each time interval a pulse of current density I is drawn off with probability p (discharge rate), with probability 1–p the cell recovers
Cut-off potential equal to 2.8V (voltage value at which the battery is considered discharged)

6. Specific Energy vs. Discharge Rate

7. Summarizing….
Pulsed current discharge outperforms constant current discharge
Battery capacity can be improved by using a bursty discharge pattern due to charge recovery effects that take place during idle periods
Given a certain value of current drawn off the battery, higher current pulses degrade battery performance even if the percentage of higher pulses is relatively small

8. Assumptions N: set of all nodes in the network
S: set of source nodes generating data traffic
D: set of destination nodes; nodes in S can direct their traffic to any of the nodes in D using other nodes as relays
Time discretized into intervals corresponding to the packet transmission time with unit duration
Nodes transmission range limited to r. Ri ( i  N) is the set of nodes whose distance from node i is  r

9. Assumptions
Energy consumed for each data transmission by the generic node i to node j:
eij={
(dij/r )4 if j  Ri
 else
Energy consumed by node j (j  Ri) to receive data from node i is 10 times lower than the transmission energy eij
Transmission energy discretized into few levels ranging from emin to emax

10. Battery Status Evolution
At the generic node i:
Energy accrue due to recovery effect modeled as a function G(li,ei) where li is the transmission rate of node i and ei is the average energy necessary to i to transmit a packet
Whenever an energy transmission eij > emin is needed, the energy loss is augmented by F(eij emin)
Thus,
bi={
bi + G(li,ei) if node i is idle
bi  [eij + F(eij emin)] else

11. The BEE Routing Algorithm
Given the generic source s and destination g, each route
rsgk is assigned the cost function
Fk=Slij  rsg [X(li)eij + pij] minirsg bi k k
lij: link between nodes i and j belonging to route rsgk
X(li): weighting function that emphasizes the energy loss of the source node. X(li)=Ali , for i=s, with A a proper constant > 1, and X(li)=1 otherwise
pij: energy penalty due to the rate-capacity effect; pij=max{0,eij ei}

12. Route Selection Fm= minrsg Fk
Whenever s has a block of data to transmit to g, the cost function is evaluated for all the possible loop-free routes and it is selected a route, rsgm, such that
Fm= minrsg Fk
If the set of routes is large, the algorithm complexity can be reduced by choosing c routes independently and uniformly at random. The cost function is computed over c routes only, and the selected path is the one among the c routes which minimizes the cost function k

13. Performance Metrics
Mean delay of the transmitted packets from the time instant when a packet is generated at the source node to the time instant when it is delivered to the destination
Network life-time: the time elapsed from the time instant when all the nodes have a fully charged battery to the time instant when the first node in the network runs out of battery

14. Simulation Scenario
Network scenario: N={0,…,14}, S={0,1,2,3,4}, D={1,14}; initial battery status=1.0; li=0.4 for i=0,...,2 and li=0.3 for i=3, 4
Results derived by simulating several network topologies randomly generated and then averaging over the obtained values
Results obtained through the BEE algorithm are compared to the performance of the MTE (Minimum Transmission Energy) scheme

15. Results - Network Life-time

16. Results - Mean Packet Delay

17. Conclusion
Routing protocol for wireless ad-hoc network that efficiently exploits battery capacity:
it selects a route with low transmission energy to avoid the rate-capacity effect
it distributes the traffic load in a manner that nodes with low battery can benefit of recovery effect
The BEE protocol can be adapted to either a slow or a fast varying network topology by changing the parameter c (and thus reducing its complexity)
Excellent performance in terms of both network life-time and packet delay